3.1. Big data analysis based on a randomly selected engineering project
More than 160 engineering cases based on the coupling system were applied in different cities of Zhejiang province, and mainly in Kaihua county, which locates in the northwest of Quzhou city. A project using this coupling system was selected and on-line monitoring equipment was added to the effluent section to investigate its long-term effect. The project locates in Xiayu village of Kaihua county, the average influent concentration of CODCr, NH4+-N and TP of rural domestic sewage in Quzhou city was 113.9 mg/L, 27.27 mg/L, and 3.00 mg/L, respectively, which basically represented the average influent water quality of Kaihua county. The main effluent indexes of the project were continuously observed for nearly 160 days, and data were automatically monitored every four hours. From Fig. 3 effluent concentrations of CODCr, NH4+-N and TP showed periodic changes, and all of the values were within their standard ranges. Table.2 showed the effluent values of the main parameters of the project, and the effluent concentrations of CODCr, NH4+-N, and TP were 37.57 ± 11.17 mg/L, 5.64 ± 1.69 mg/L, and 0.82 ± 0.16 mg/L, respectively. On the whole, Fig. 3 showed a relatively stable and good effluent water quality, which met the first level of "Standards for Discharge of Water Pollutants from Rural Domestic Sewage Treatment Facilities" (DB33/973–2015).
It also could be found from Table.2 that the flow rate of the coupling system was 1079.36 ± 37.79 m3/h (ranges from 25.0 m3/d to 26.8 m3/d) during the monitoring period, which approaching the designed parameter (27.0 m3/d). The emergence of the situation that the effluent flow rate was lower than the design flow rate may be related to the gap between the design and the actual situation, leakage or impervious measures in constructed wetland or sewage pipe network, plant transpiration, and other problems.
Figure 4 showed correlations of the main effluent water quality indexes with each other. As shown in Fig. 4(a), the concentrations of effluent CODCr had a strong positive linear correlation with the concentrations of effluent NH4+-N (R2 = 0.9297), and increased with the increase of effluent TP concentrations in Fig. 4(c) (R2 = 0.6957). While in Fig. 4(b), the concentrations of effluent TP showed a strong correlation with effluent NH4+-N concentrations (R2 = 0.9329). The results showed that there was a strong correlation between different effluent water quality indexes, and if the sample capacity was large enough, the data would be much more accurate. This founding might have very important guiding significance for the future monitoring methods of effluent water quality of treatment facility and the reduction of monitoring management cost, as well as to ensure the main effluent water indexes meet the standard by a combination of big data analysis and AI technology in ways of early warning and feedback control. In this study, the effluent water index of domestic sewage had stronger regularity, and it was found that the effluent also showed a strong correlation when analyzing other treatment objects such as surface water and other process technologies. This might be the result of complex physical, chemical and biological reactions occurred in different facilities, making it more regular than untreated water.
Few relevant studies and reports were found about effluent water index than the influent in literature. Sun (Sun et al., 2014) found that the indicators including BOD5, CODCr, TN and TP had significant simple linear relationships. And the correlation coefficient of BOD5 and CODCr, BOD5 and TN, BOD5 and TP, TN, and TP was 0.968, 0.995, 0.823, 0.817, respectively. Besides, some scholars were looking for new methods that can be used for rapid or low-cost simultaneous monitoring of multiple indicators. Wang (Wang et al., 2019) tried to obtain a satisfactory prediction accuracy modelling approach to identify some hard-to-measure variables like COD and TP by selecting flowrate. Lotfi (Lotfi et al., 2019) presented a methodology to model the quality parameters of effluent wastewater and found that the methodology could accurately predict BOD in wastewater. Narendra (Narendra et al., 2019) developed a feedforward artificial neural network model to predict effluent quality parameters through influent characteristics like pH, TSS, BOD, COD, TKN, and TP etc.
Figure 5(a)-(d) was the correlations between pH and effluent CODCr, NH4+-N, TP, C/N.As shown in Fig. 5, C/N showed a stronger correlation with pH (R2 = 0.6441), while the concentrations of effluent NH4+-N and effluent TP had weak correlation with pH, and the correlation coefficient of which was R2 = 0.3348 and R2 = 0.4834, respectively. It could be seen from Fig. 5 (d) that the pH decreased with the increase of C/N when the effluent C/N was less than 8.5, thus a method might be derived to fast monitor and control the effluent quality indexes. Moreover, if the amount of data sample was big enough, the effluent C/N could be used as a more accurate indicator, and the effluent C/N could be indirectly reflected by the change of easily monitored pH value, making the influent C/N could be changed quickly according to the effluent requirements, or the effluent water quality could be improved by fast feedback and the changing of HRT.
Although the project based on the coupling system achieved good and stable effluent water quality, it was only a randomly selected case, and could not represent the entire application effects of the coupling system. Therefore, it was necessary to study more cases and to summarize the actual project effluent effects of this coupling system.
3.2. Study on the pollutant removal in 12 different randomly selected engineering cases
As shown in Table.1, in order to better understand the actual engineering effect and existing problems of the technology more comprehensively, 12 engineering cases with different scales (3-100 m3/d) were randomly selected for the comparative study of pollutant removal effect.
According to investigation conducted our group in year 2015, the average concentration of CODCr, NH4+-N and TP in the influent of Zhejiang Province (except Ningbo) was 186.35 mg/L, 30.33 mg/L and 4.06 mg/L, respectively, and the average C/N of Zhejiang Province was 6.14, which was beneficial to biological nitrogen and phosphorus removal. The 12 engineering cases in this study mainly located in Quzhou City and Wenzhou city, where contains many mountainous areas. And its average influent concentration of CODCr, NH4+-N and TP was 171.73 mg/L, 22.14 mg/L and 3.20 mg/L respectively, which was lower than the average values of Zhejiang Province. Besides, there were significant differences among these cities (P < 0.05). Result showed that the topography and economic development level of mountainous areas had greater impact on the influent water quality.
The data of the study case in Fig. 6 showed that the effluent concentrations of CODCr, NH4+-N and TP in the 12 selected projects were 31.23 mg/L, 9.27 mg/L and 0.97 mg/L, respectively, and the removal efficiencys were 81.81%, 58.13% and 69.69%, respectively. Compared with other research results (Lu et al., 2015; Lu et al., 2016), these data were not ideal. However, results of this study reflected the effect of practical engineering application, while other studies were on the bench-scale or pilot-scale. Result of the investigation taken in 2015 found that the average CODCr, NH4+-N and TP effluent concentration of Zhejiang Province was 35.62 mg/L, 10.90 mg/L and 1.55 mg/L, respectively. Which means the main effluent indexes of the coupling system were better than that of in Zhejiang Province, and had relatively good treatment effect. Zhejiang rural domestic sewage treatment facilities were required to implemented the Rural Domestic Sewage Treatment Facilities Water Pollutant Discharge Standard (DB33/973–2015) since 2015, in which the effluent CODCr concentration specified in the first level is 60 mg/L, while the NH4+-N concentration is 15 mg/L, and the TP concentration is 2 mg/L. The coupling system could mostly reach the requirement based on the data from previous practical engineering applications but the exceptions of 9# facility with the effluent CODCr concentration of 69.08 mg/L, 4#, 6#, 8# and 9# facility with effluent NH4+-N concentration, and 4#, 6# facility with effluent TP concentration all exceeded the first level. It should be noted that for the 6# facility with continuous monitoring equipment introduced above, its NH4+- N and TP exceeding standards in this sampling test, which might be related to factors such as the operation phase of the system and the different monitoring time, Therefore, the effluent water quality and stability of practical engineering cases need to be further improved, and the technical process need to be optimized to achieve a stable effluent water quality.
3.3. Comparative study on the treatment of CODCr, NH4+-N and TP in demonstration based on the optimized coupling system.
As mentioned above, a demonstration comparison system based on this coupling system was conducted (Fig. 2) in order to further improve the sewage treatment effect and operation stability of this technology. Figure 7 showed the long operation and removal of CODCr, NH4+-N and TP by the two contrast coupling systems. The system-b represents the enhanced system (b) while the system-a stands for the original system (a).
As can be seen from Fig. 7, the variation of CODCr of influent water was relatively large, which from 350 mg/L at the beginning to below 50 mg/L, then suddenly rose to about 400 mg/L on the 30th day; And gradually declined after the 45th day, lower than 100 mg/L. Compared with the original system (a), the enhanced system (b) had lower effluent CODCr concentration. Especially during the period when the influent CODCr concentration fluctuates sharply. For example, the effluent CODCr concentration of the enhanced system (b) was lower and more stable from the 20th to the 70th day, and the maximum difference of removal efficiency was more than 50%. Figure 7 also showed that the temperature increases with time, and the CODCr removal efficiencys of the two coupled contrast systems were increasing. On the whole, as shown in Table.3, the average CODCr concentration in the influent was 152.81 ± 130.33 mg/L, while the CODCr concentrations in the effluent of the original system (a) and the enhanced system (b) were 23.63 ± 13.89 mg/L and 20.22 ± 13.37 mg/L, respectively. Vymazal (Vymazal ., 2019) evaluated 17 HFCWs operated for more than 20 years and revealed that the removal efficiency of these systems amounted to 82.9% for COD, while in this study, the average removal efficiencys of CODCr in the original system (a) and the enhanced system (b) was 77.00% and 81.33%, respectively. The removal efficiency of CODCr in the enhanced system (b) was higher than that in the original system (a), which rose by 4.33%.
The fluctuation of influent NH4+-N was also large during the study period. Figure 7 showed that the concentration range of influent NH4+-N was 11.10–58.00 mg/L, and the effluent NH4+-N concentration of the two coupled systems is relatively low and stable in the first 30 days. With the continuous increase of influent NH4+-N concentration (up to 58.00 mg/L) from 30 to 45 days, the effluent NH4+-N concentration of the two coupled systems showed an upward trend, and the effluent NH4+-N concentration of the enhanced system (b) was significantly lower than that of the original system (a) (P < 0.05). On the 53rd day, the removal efficiency of NH4+-N in the original system (a) reached the lowest 27.08%, while that in the enhanced system (b) was close to 80.00%, showing strong stability. In summary, the average NH4+-N concentration of the original system (a) and the enhanced system (b) was 5.93 ± 5.29 mg/L and 2.70 ± 2.49 mg/L, respectively. The average removal efficiency was 81.11% and 90.22%, respectively. That is to say, the removal efficiency of NH4+-N by the enhanced system (b) was 9.11% higher than that of the original system (a). Yasinta (Yasinta et al., 2020) found the aerated saturated vertical up-flow constructed wetland achieves a high (more than 85%) removal of NH4+-N, while the new coupling system had a much higher removal efficiency with nonexternal power assisted aeration process. Different substrates exhibit different adsorption characteristics. Khalifa (Khalifa et al., 2020) found in their research that some traditional used media could boost the pollutants removal capacity of polystyrene foam. Results showed that the removal of NH4+-N increased from 66 to78%, while a better removal effect was obtained in this study without adding polystyrene foam.
It could be seen from Fig. 7 that the removal effect of TP by the two coupled systems was also very different. Although the influent TP concentration varied considerably, most of the effluent concentrations of the enhanced system (b) were less than 1.00 mg/L. On the whole, the effluent TP concentration of the enhanced system (b) was significantly lower than that of the original system (a) (P < 0.05), which maintained a relatively stable TP removal efficiency. The average TP concentration of the original system (a) and the enhanced system (b) was 1.66 ± 0.73 mg/L and 1.09 ± 0.62 mg/L, respectively. The average removal efficiency was 61.67% and 75.44% respectively. That is to say, the removal efficiency of TP by the enhanced system (b) was 13.77% higher than that of the original system (a).
Table.3 shown the 6 months continuous monitor data of effluent water indexes derived from the officially authorized testing agency. The effluent monitor results of the two comparison systems showed that the effluent CODCr, NH4+-N and TP of the demonstration project reached the first level of "Standards for Discharge of Water Pollutants from Rural Domestic Sewage Treatment Facilities" (DB33/973–2015). It could be found from Fig. 7 and Table.3 that the comparison system showed better removal efficiency compared with the other one, although the influent fluctuated unregularly. And the comparison system had great superiority on the removal of NH4+-N and TP than that of the original system.